>I still can't ask my computer to do anything in free form unless it has been specifically programmed to do it: "find the fastest public transport route from place A to place B, but disregard all busses unless they cut down the transit time by more then 20 minutes".

This feels a bit like goalpost moving as the basic feature did become popularized in the past 10 years, but the complexity is limited. You can certainly ask most smart devices for the fastest route from A to B and get it. Most people use digital assistants that take plain-text voice queries, like Alexa or Google. At my home we use it often to set timers or ask questions when cooking, like "what temperature should ground beef be cooked to" or others and it usually just works.

It's also somewhat more difficult to use AI for consumer-focused products compared to industry-focused products. AI systems like AlphaFold are currently in the background creating significant speedups in biological research and development. It is also known that multiple large hardware companies are also using AI to accelerate computer chip design.

>Space travel (do I even need to explain this one?)

The first successful Falcon 9 landing was at the end of 2015. Since then there have been individual Falcon 9 boosters which have been reused 14 times. Space tourism is slowly becoming a thing with Blue Origin's New Shephard having completed six crewed flights, the first one being in July 2021.

Talking about things like self-driving, nuclear fusion and androids is sort of a red herring. Just because a given task is more difficult than expected, is not evidence that we've reached a plateau. It just means the thing was harder than originally thought. There are various futuristic technologies which have come to market and started to mature. Electric vehicles being a big and obvious one, now with every major automaker producing at least one mainstream EV and having concrete plans to do many more (and in some cases convert their entire fleets to electric).